Skip to content

Large Model Visualization

For large neural network models, the profiling data can become very large. To address this, Perfetto's official web visualizer page imposes a limit on memory of over 2GB (this size limit is not related to the profiled Protocol Buffer file size). To work with such large models and logs, this section provides instructions on how to host Perfetto locally, allowing you to visualize and analyze profiling results on your own server.

For more detailed information on how to handle large traces, please refer to the official Perfetto documentation.

1. Download and Execute Trace Processor

Perfetto’s GitHub repository provides an executable file that allows you to host a server and load profiled files in local memory. Below is the method to execute trace_processor_shell.exe in Windows OS or trace_processor in Mac & Linux OS. When you execute the files, the default port number is 9001.

  • On Windows (Powershell) 

    1
2
3
4
    > wget https://github.com/google/perfetto/releases/download/v49.0/windows-amd64.zip
> unzip windows-amd64.zip
> cd windows-amd64\windows-amd64\
> ./trace_processor_shell.exe --httpd /path/to/rbln_xxxx.pb

  • On Mac & Linux (bash) 

    1
2
3
    $ curl -LO https://get.perfetto.dev/trace_processor
$ chmod +x ./trace_processor
$ ./trace_processor --httpd /path/to/rbln_xxxx.pb

2. Go to Viewer

Note

This step may not function correctly on the Safari browser in macOS. For optimal performance, it is recommended to use the Chrome browser.

  • Web Perfetto detects version mismatch with the local tracer. Click “Open https://ui.perfetto.dev/v49.0-33a4fd078/” to access the hosted local tracer.

  • Select "YES, use loaded trace" to visualize Protocol Buffer file which profiled your model.

3. Final View

You can confirm that the local tracer is successfully visualized through the display "RPC@127.0.0.1:9001" or “rbln_xxxx.pb (file 1)” on the left.